Perception of Leitmotives in Richard Wagner's Der Ring des Nibelungen

Abstract

The music of Richard Wagner tends to generate very diverse judgments indicative of the complex relationship between listeners and the sophisticated musical structures in Wagner's music. This paper presents findings from two listening experiments using the music from Wagner's Der Ring des Nibelungen that explores musical as well as individual listener parameters to better understand how listeners are able to hear leitmotives, a compositional device closely associated with Wagner's music. Results confirm findings from a previous experiment showing that specific expertise with Wagner's music can account for a greater portion of the variance in an individual's ability to recognize and remember musical material compared to measures of generic musical training. Results also explore how acoustical distance of the leitmotives affects memory recognition using a chroma similarity measure. In addition, we show how characteristics of the compositional structure of the leitmotives contributes to their salience and memorability. A final model is then presented that accounts for the aforementioned individual differences factors, as well as parameters of musical surface and structure. Our results suggest that that future work in music perception may consider both individual differences variables beyond musical training, as well as symbolic features and audio commonly used in music information retrieval in order to build robust models of musical perception and cognition.

Examining musical sophistication: A replication and theoretical commentary on the Goldsmiths Musical Sophistication Index

“…Recently the test has been modified and adapted using more sophisticated statistical modeling, resulting in shorter versions of the test, which reportedly have the same construct validity as the original (Harrison, Collins, & Müllensiefen, 2017). In its complete form, the Gold-MSI takes about 30 minutes to complete, although some researchers elect to only use relevant subcomponents such as solely the Musical Training sub-scale (Baker & Müllensiefen, 2017;Farrugia, Jakubowski, Cusack, & Stewart, 2015). The tool has proved especially useful in studies where it has been combined with other measures such as measures of personality (Greenberg et al, 2015) and socio-economic status (Müllensiefen et al, 2014), and has been used as a tool to measure aspects of development (Müllensiefen, Harrison, Caprini, & Fancourt, 2015).…”

A Computational Model of Melodic Dictation

“…Features derived from the FANTASTIC toolbox have been successful in predicting a range of musical perception behavior (Baker & Müllensiefen, 2017;Harrison et al, 2017;Jakubowski et al, 2017;Kopiez & Müllensiefen, 2011) and thus serve as a useful metric from which PERCOREMA can be compared. In order to make this benchmark comparison, PECOREMA's performance is visualized against five features from the toolbox.…”

Frequently Occurring Melodic Patterns are Easier to Recall

“…For example, there currently exists rationale for using computational measures to model aspects of musical cognition (Pearce, 2018) when memory is conceptualized as compressibility (Eerola et al, 2009) using information content frameworks. For example, Pearce and Müllensiefen found that measures of compressibility can be used as predictors of musical similarity (Pearce & Müllensiefen, 2017) and work using symbolic summary features has additionally been successful at modeling musical memory using computational measures of complexity (Baker & Müllensiefen, 2017; J. C. Bartlett & Dowling, 1980;Cuddy & Cohen, 1976;Cuddy & Lyons, 1981;Harrison et al, 2017).…”

“…For example, it would be difficult to attempt to change an estimation of tonalness by adding in non-diatonic notes to a melody without altering the pitch or interval entropy calculations. Some researchers have attempted to side-step this problem by using data reductive techniques such as principal component analysis in order to distill features from the melodic data to a single complexity score (Baker & Müllensiefen, 2017;Harrison et al, 2017), but given the degree of predictive ability from measures here related to information content, adding a data reductive model here might make things more difficult to interpret for future work.…”